Deformation Measurement Method Of Painted Beams In Ancient Buildings Based On Improved Digital Image Correlation

Chinese ancient building is the organic combination of exquisite construction technology and excellent traditional culture,and it is the precious wealth of the Chinese nation.In ancient building structures,wooden beams are the main flexural load-bearing components,which bear long-term vertical loads such as snow from roofs,floors and crowds.Wood will inevitably aging,which will reduce the stiffness and increase the deformation of wood beams in the course of using for hundreds of years or even thousands of years.Therefore,the structural safety state of wooden beams in ancient buildings can be mastered by deformation monitoring.However,traditional deformation measurement methods often need to install monitoring sensors on wood beams,which will inevitably cause damage to wood structures.Digital close-range photogrammetry is a new non-contact deformation measurement technology,and the algorithm for calculating pixel change is one of its core technologies.It uses high-performance digital camera to obtain the image of the object under test,and then calculates the change of the pixels in the image to achieve deformation measurement.In this paper,an adaptive search double-precision gradient DIC algorithm is proposed for the deformation calculation of painted wood beams based on the principle of digital image correlation and the gray distribution characteristics of the painted beam wooden beams which provide technical support for non-contact deformation measurement of ancient architectural painted structures.The main research contents and conclusions of this paper are as follows:(1)A bidirectional adaptive image subset selection method is proposed based on the color pattern features and the gray gradient square criterion according to Sum of Square of Subset Intensity Gradients(SSSIG)criterion.By independently changing the dimensions of the two axes of the subset,the optimal subset size can be searched adaptively,and the advantages of this method in choosing the best shape and size of subset are verified by color drawing.The computational efficiency is improved by 14% without affecting the computational accuracy when the subset calculated by this method is used in digital image correlation method.(2)Part Exhaustive-Adaptive Rood Pattern Search method is proposed to solve pixel displacement aiming at the problem of uneven gray scale of wooden beam color drawing in ancient buildings,which results in poor recognition effect and low calculation efficiency of traditional whole-pixel algorithm.Predict the search radius of the first search point,search exhaustively within the radius,and then use the adaptive path search method to obtain the corresponding pixel displacement.The first calculation point is searched exhaustively within the estimated search radius,then Adaptive Rood Pattern Search method is used to obtain the integer pixel displacement of adjacent computing points.It can be seen from the results of bending tests of color-painted beams with common patterns that the proposed method improves the accuracy of displacement calculation by 3.03%,3.39% and 11.1% respectively,while improving the computational efficiency by 71.6%,51.2% and 59.5% compared with the traditional coarse-fine search algorithm,three-step search algorithm and diamond algorithm.(3)Double-Precision-Gradient method is proposed.The sub-pixel displacement between two adjacent integral pixels is obtained by use Gradient Method that using these two integer pixels as initial values respectively.Then,according to the standard deviation of image noise,the non-linear weight coefficients of the two calculation results are allocated and summed to eliminate the errors introduced by the gradient method in the first-order expansion process.The simulation example of speckle pattern and the actual test results of wood beam color drawing show that the double precision gradient method can not only effectively eliminate the errors caused by the calculation of sub-pixel displacement by the initial point of a single integral pixel,but also has good noise resistance.(4)A high-precision method for calculating AD-DIC displacement of wooden beams in ancient buildings is established by combining PE-ARPS and DPG algorithm.Three kinds of color drawing of wooden beams of ancient buildings are used to verify the applicability of the proposed method through bending tests,and different lighting conditions are set to simulate the light effects in the measured environment.The experimental results show that the AD-DIC method improves the calculation accuracy by 2.68% and the calculation stability by 75.5% compared with the traditional method without obvious reduction in calculation efficiency.The recognition accuracy is 99.22% under 6.43%-12.32% illumination,which is 51.2% higher than that of traditional methods.